DC-DC Conversion Poweringfor the CMS Tracker at SLHC

Lutz Feld, Rüdiger Jussen, Waclaw Karpinski,Katja Klein, Jennifer Merz, Jan Sammet

RWTH Aachen University

12th Vienna Conference on Instrumentation - VCI 2010

17. 2. 2010

Lutz Feld 2

Overview

• Motivation and Strategy

• Powering Scheme with DC-DC Converters

• Problems and Challenges … and how to address them

• R&D with commercial ASICs

• R&D with custom radiation hard ASICs

• Conclusions

17. 2. 2010

Lutz Feld 3

Motivation

17. 2. 2010

CMS Tracker Endcap TEC+Tracker before insertion into CMS

• current CMS and ATLAS trackers burn 50% of their power in supply cablesCMS: front end 33kW, cables 34kW, total current 15kA

Lutz Feld 4

Motivation

• pixel and strip tracker upgrades will likely need more power than current systems :– more channels (more pixel layers, higher granularity)– more functionality (contribution to L1 trigger)

• …at a lower voltage (smaller ASIC feature size)

currents will increase substantially cable losses rise quadratically: Ploss=R·I2

increase power cable cross-section … not possible

increase copper cross-section inside tracker … not desirable

need an improved powering scheme

17. 2. 2010

CMS Tracker cable routing

cablescooling

CMS Strip Tracker material budget

Lutz Feld 5

• P = U·I = (rU) ·(I/r)

supply power at higher voltage i.e. lower current

cable loss Ploss=R·I2 is reduced by factor r2!

Strategy

17. 2. 2010

DC-DC Powering

PSmoduleDC-DC

moduleDC-DC

Conventional Powering

PSmodule

module

Serial Powering

PS module moduleshunt shunt

Lutz Feld 6

Choice of Powering Scheme• both new powering schemes seem feasible

• DC-DC conversion– system design is simpler and closer to current systems, less interdependencies– only one new component– switching noise is a concern

• serial powering– no switching– each module has different ground potential (AC communication, …)– modules in a chain are coupled, need safety features (shunts, …)

• little difference in efficiency and material budget

• CMS Tracker has chosen DC-DC conversion as baseline– mainly because changes to the system are minimal and factorizable– serial powering should be kept as a back-up

17. 2. 2010

Lutz Feld (RWTH Aachen) 7

DC-DC Converters• many different designs

• inductor-based converters– current capacity up to several amps– ferrite cores saturate in 4 Tesla field

need to use air-core coils

• capacitor based converters– are limited in current to few 100 mA at most– no inductors needed, can be very compact

and may be even included in FE ASIC (external capacitors)

• piezoelectric transformers

• all need rad-hard “high voltage” transistorsas switches

• efficiency typically 70 – 90 %

• switching noise is a concern19.10.2008

Inductor Based “Buck Converter”

Capacitor Based “Charge Pump” [10]

Lutz Feld 8

output voltage is regulated via feed-back loop by adjustment of

duty cycle :

for lossless converter

on

out in

tD

TV D V

operation principle

DC-DC Buck Converters

17. 2. 2010

19mm

12mm

+-

Vin

Vo

VI

VFS2

D2

IND

BTS

TR

+-

driver

Cout

L

CBTSR

VDD (for control)

VSS

SUB

2.5V

1.8V 5V

Ena

ble

freq

Rf

Enable delay

R delay

VF VI

VDD2 (for drivers)

Sawtooth generator

En_BNGPVref

SW1

SW2

SW

1

SW

2

Vref

Vref

IPTA

TENABLE

VR

AM

P

+-

Vin

Vo

VIVFS2

D2

IND

BTSTR +-

drive

r

C out

L

C BTS

R

VDD

(for c

ontro

l)

VSS

SUB

2.5V

1.8V

5V

Enable freq

Rf

Enab

le de

lay

R de

lay

VF

VI

VDD2

(for

drive

rs)

Sawt

ooth

ge

nera

tor

En_B

NGP

VrefSW

1

SW2

SW1

SW2

Vref

Vref

IPTAT

ENAB

LE

VRAMP

inV outVont

2...10

  1...4

60 90%

1

conversion ration

switching frequency MHz

efficiency

current ripple at inductor

in

out

out

in

outL

VrV

fPP

V DI

L f

main parameters

2~

resistive loss in inductorresistive loss in MOSFETs

switching loss in MOSFETs

input and output capacitor ESR lossesin outV I f

main loss mechanisms

Lutz Feld 9

Challenges with DC-DC Buck Converters

• coupling of switching noise into detector modules– conductive– radiative

• air core inductors required due to 4 Tesla field– bulky– radiates noise

• radiation hard ASIC needs to switch rather high voltages (~10V)– need special high voltage process

• useful only if efficiency is high

• total material budget should be reduced despite addition of converters

• small size required for integration close to detector modules

17. 2. 2010

Lutz Feld 10

R&D with commercial ASICs• buck converters with commercial non-radiation-hard chips• used to optimize for low mass, low space, low noise, and for studies in system test

17. 2. 2010

Chip: Enpirion EQ5382DVin = 2.4-5.5V(rec.)/7.0V(max.)Iout 0.8Afs 4MHz

PCB:2 copper layers a 35mFR4, 200µmV = 2.3cm2 x 10mmm = 1.0g

Input/output filters

Air-core inductor:Custom-made toroid, 6mmL = 200nH or 600nH

19mm

12mm

AC2

Lutz Feld 11

R&D with radiation hard custom ASICs

• radiation levels at 22cm & 3000fb-1: fluence ~ 1015/cm2 , dose ~ 1MGy• two candidate technologies identified by F. Faccio, CERN [contrib. to TWEPP 09]• ASICs developed by S. Michelis, CERN [contrib. to TWEPP 09]

17. 2. 2010

+-

Vin

Vo

VI

VFS2

D2

IND

BTS

TR

+-

driver

Cout

L

CBTSR

VDD (for control)

VSS

SUB

2.5V

1.8V 5V

Ena

ble

freq

Rf

Enable delay

R delay

VF VI

VDD2 (for drivers)

Sawtooth generator

En_BNGPVref

SW1

SW2

SW

1

SW

2

Vref

Vref

IPTA

T

ENABLE

VR

AM

P

Chip: AMIS2 by CERNVIN = 3 - 12VIOUT < 3AVOUT = 1.2V, 2.5V or 3.3VfS ≈ 1.3MHz or programmable betw. 600kHz...4MHz

PCB:2 copper layers a 35mFR4 1mmA = 18mm x 25mm for QFN32

Input and output π-filtersL = 12.1nH, C = 22µF

Air-core toroid:Custom-made toroid, 6mm,height = 7mm, L = 600nH, RDC = 80mΩ

25mm

18m

m

m = 2.7gAMIS2

Lutz Feld 12

Bx BzBy

BxBy

Bz

Radiated Switching Noise

• current ripple in inductor radiates electromagnetic fields (near field)• even a toroid radiates, dipole characteristics (confirmed by simulation)• shielding with 30µm aluminium works very well and adds little material

17. 2. 2010

Whole PCB shielded

Lutz Feld 13

AC2 Diff. ModeVin=5.5V, Vout=1.3V

AC2 Diff. ModeVin=5.5V, Vout=1.3V

with Pi filter

Conducted Switching Noise

impact on detector system depends on its frequency dependent susceptibility

17. 2. 2010

SpectrumAnalyzer

Load

LISN = Line ImpedanceStabilization Network

Lutz Feld 14

System Test with CMS Strip Modules

17. 2. 2010

6.16.4 6.3 6.2

Motherboard

Ring 6 modules

• SLHC readout chips and module prototypes not available before 2011• use current CMS tracker hardware for system test of DC-DC converters

TEC petal

• two DC-DC converters per module• integrated via additional adapter• Vin from external power supply

APV25 readout chip:- 0.25 µm CMOS -128 channels- analogue readout- per channel: pre-amp., CR-RC shaper, pipeline- on-chip common mode subtraction- = 50ns - 1.25V & 2.50V supply- I250 = 0.12A, I125 = 0.06A

Lutz Feld 15

System Test with CMS Strip Modules

17. 2. 2010

Conventional powering

Conventional powering

• Raw noise: RMS of fluctuation around pedestal value• Edge channels are particularly sensitive (due to on-chip common mode subtraction)• Large increase with previous generation of boards (AC1), in particular on edge strips; both conductive (ripple) and radiative (inductor) contributions (TWEPP08)

{--- Conventional powering--- DC-DC converter (AC1, 2008)

1 APV = 128 strips

--- Conventional powering--- DC-DC converter (AC1, 2008)--- Conventional powering--- DC-DC converter (AC1, 2008)--- DC-DC converter (AC2, 2009)

Zoom onto edge channels

--- Conventional powering--- DC-DC converter (AC1, 2008)--- DC-DC converter (AC2, 2009)

2 21 512N N noise of edge strips

Lutz Feld 16

Results with Commercial Converters

• 4 MHz, conversion ratio up to 5.6, output current ~0.5 A• no shielding• simple Pi filter suppresses conductive noise efficiently

17. 2. 2010

DC-DC powering without any noise increase is feasible

Lutz Feld 17

Results with radiation-hard ASIC (AMIS2)

17. 2. 2010

• 1 - 4 MHz• conversion ratio up to 8• output current 0.5 A• no shielding

• at operating conditions similar tocommercial converters (4MHz, …)– rad-hard ASIC is similarly quiet– efficiency is reduced due to

switching loss

• at 1MHz switching frequency– better efficiency– higher ripple leads to conductive

and radiative noise– can be controlled by filtering,

shielding and/or distance to module

Lutz Feld 18

Understanding the Noise

17. 2. 2010

Long-term reproducibility

AC1AC2 mounting

AC2-Stand.C AC2-Rev.C AC2-IDC

No converter Mini Toroid, 600nH Tiny Toroid, 200nH

• noise caused by current ripple in inductor

• low-ESL in capacitors improve filtering

1outL

V DI

L f

D decreases

f increases

L increases

Lutz Feld 19

Efficiency

• ohmic losses (inductor, Ron, …) increase with output current

• switching losses increase with frequency and input voltage

• efficiencies of 60 … 80 % reached• still room for improvement

17. 2. 2010

AMIS2 chip, Vout=1.2 V, 1.3 MHzEnpirion chip, Vout=1.3 V, 4 MHz

Lutz Feld 20

Implementation into a Pixel/Strip System

• CMS plans to employ DC-DC converters

in Phase I Pixel upgrade

– converters will sit on pixel support tube

at |eta|~4

material budget, size and radiated noise

much less critical

– prototyping is underway

• Phase II upgrade of full tracker will very likely

include trigger layers with high power density

– DC-DC conversion is essential

– converters should be installed next to modules

– more R&D needed

17. 2. 2010

Pixel Barrel

DC-DC converters

Lutz Feld 21

Material Budget

• case study for current CMS Tracker End-Caps

– full Geant4 simulation

– typical converter: 10% of module radiation length

– one converter per module

– assume conversion ratio of 8 and 80% efficiency

– re-calculate cable/trace cross sections

save 30.9% on ‘electronics & cables’

save 8% of total material budget

• similar savings for serial powering (7.5%)

17. 2. 2010

Electronics & cables Old layoutDC-DC conv.- 30.9%

Lutz Feld 22

Summary

• DC-DC powering can supply the large currents at low voltage which are needed for pixel and strip tracker upgrades

• simulations show that total material budget can be reduced by employing a DC-DC conversion powering scheme

• radiative and conductive noise is an issue, but can be controlled by filtering, board design, shielding, distance, conversion ratio, switching frequency

• non-rad hard converter prototype with commercial components in hand with 75-85% efficiency, I=0.8 A, conversion ration ~4, low mass (~10% of strip module), small size (19x12 mm2), magnetic field tolerance and no significant impact on system noise

• first rad-hard prototypes under test with promising results

17. 2. 2010

Lutz Feld 23

back - up

17. 2. 2010

Lutz Feld 24

Irradiation Results for Transistors[Federico Faccio, TWEPP 2009]

17. 2. 2010

0.0E+00

5.0E-04

1.0E-03

1.5E-03

2.0E-03

2.5E-03

3.0E-03

3.5E-03

4.0E-03

4.5E-03

5.0E-03

0 2 4 6 8 10 12 14 16Vds (V)

Ids

(A)

prerad1e15 p/cm22e15 p/cm28e15 p/cm21e16 p/cm2

0.0E+00

2.0E-04

4.0E-04

6.0E-04

8.0E-04

1.0E-03

1.2E-03

1.4E-03

1.6E-03

1.8E-03

0 2 4 6 8 10 12 14Vds (V)

Ids

(A)

prerad1e15 p/cm22e15 p/cm26e15 p/cm28e15 p/cm21e16 p/cm2

IHP NMOS IHP PMOS

0%

50%

100%

150%

200%

250%

300%

1.E+13 1.E+14 1.E+15 1.E+16

Proton fluence (p/cm2)

Ron

incr

ease

(%)

A, 0.35umB, 0.25umC, 0.18um

D, 0.18umE, 0.13um

0%

50%

100%

150%

200%

250%

300%

1.E+13 1.E+14 1.E+15 1.E+16

Proton fluence (p/cm2)

Ron

incr

ease

(%)

A, 0.35um

B, 0.25um

C, 0.18um

D, 0.18um

E, 0.13um

NMOS PMOS

[http://indico.cern.ch/contributionDisplay.py?contribId=163&sessionId=25&confId=49682]

IHP

3E14 3E14

AMISIHP

AMIS

Lutz Feld 25

AMIS2 [Stefano Michelis, CERN] •VIN +3.3V to +12V •programmable up to 2.5MHz•internal voltage reference •integrated feedback loop withbandwidth of 20Khz•different Vout can be set:1.2V, 1.8V, 2.5V, 3V, 5Vother values by ext. resistor•lateral HV transistors are usedas power switches•package QFN32

•efficiency measured 70…80%•tested up to 300 Mrad = 3000 kGy withonly 2% efficiency loss (after annealing)•noise spectra look good (=low)

•see Stefano‘s talk at TWEPP 2009: http://indico.cern.ch/contributionDisplay.py?contribId=97&sessionId=42&confId=49682

17. 2. 2010

+-

Vin

Vo

VI

VFS2

D2

IND

BTS

TR

+-

driver

Cout

L

CBTSR

VDD (for control)

VSS

SUB

2.5V

1.8V 5V

Ena

ble

freq

Rf

Enable delay

R delay

VF VI

VDD2 (for drivers)

Sawtooth generator

En_BNGPVref

SW1

SW2

SW

1

SW

2

Vref

Vref

IPTA

T

ENABLE

VR

AM

P

Aachen Converter Board

Lutz Feld 26

Leakage (sat)

1.E-12

1.E-11

1.E-10

1.E-09

1.E-08

1.E-07

1.E-06

1.E-05

1.E-04

1.E-03

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08

TID (rad)

Leak

age

(A) A3

A2

A1

C1

Vth (linear)

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

1.E+02 1.E+04 1.E+06 1.E+08

TID (rad)

Vth

(V) A3

A2

A1

C1

AMIS 2 Irradiation Results [Stefano Michelis, CERN]

• X-ray radiation tests shows a decrease of the efficiency mostly due to the radiation induced leakage current , compensated by the threshold voltage shift

17. 2. 2010

Pre rad Annealing7 days

Annealing3 days

Lutz Feld 27

Noise Susceptibility of APV25 Strip Modules

17. 2. 2010

• Goal: identify particularly critical bandwidth(s) for converter switching frequency• Bulk current injection (BCI) method used• A noise current of 70dBA (Ieff = 3.16mA) is injected into the power lines

Differential Mode (DM) and Common Mode (CM) on 2.5V and 1.25V

CMS Module

Lutz Feld 28

Noise Susceptibility of APV25 Strip Modules

17. 2. 2010

Step width: 0.1MHz for 100kHz-10MHz,1.0MHz for 10MHz-30MHz,2.5MHz for 30MHz-100MHz

• Peak at 6-8MHz, well above future switching frequency (3.2MHz exp. from shaping time)• Higher susceptibility for DM and 1.25V = pre-amplifier reference voltage• Set-up will be valuable to characterize future module prototypes

Noisedistributions

Edge stripnoise

Plot vs. f

Lutz Feld 29

EMI Susceptibility of Strip Modules• noise injected at 4 MHz• sender is scanning 10 mm above module• color code shows average module noise

17. 2. 2010

y [m

m]

ADC counts

y [m

m]

ADC countsE-Field injectedB-Field injected

ADC countsx [mm] x [mm] ADC counts